ABSTRACT There is a growing interest in dietary patterns that capture the overall quality of diet as well as its constituent foods and nutrients. Commonly used dietary patterns are a priori diet score/index based on a set of dietary recommendations for a healthy diet (e.g., Mediterranean diet, Healthy Eating Index) or data-driven dietary patterns (e.g., prudent diet, western diet). Numerous studies have shown that those dietary patterns were related to the risk of chronic diseases such as heart disease, diabetes, and cancer. However, none of these dietary patterns incorporates eating behavior such as when we eat (i.e., eating time) and how often we eat (i.e. eating frequency) during a day. Since the amount of foods and nutrients consumed at one eating occasion influences the food consumption at the subsequent eating occasion and overall intake of the day, eating time and frequency are integral parts of dietary patterns. Furthermore, several lines of evidence consistently suggest that eating time and frequency as well as a meal composition play roles in body weight regulation and metabolic health and also regulate circadian rhythms, all of which may lead to metabolic dysfunctions and ultimately chronic diseases. Given a clear need to expand the dietary patterns framework and close a gap in dietary patterns methodological work, we propose to 1) develop a ?temporal? dietary patterns based on temporal distribution of eating time and frequency during a day; and 2) evaluate if the identified temporal dietary patterns are associated with i) overall diet quality and nutrient intakes, ii) adiposity (e.g., BMI, waist circumference), and iii) metabolic biomarkers (e.g., insulin, HOMA-IR, LDL-cholesterol, c-reactive protein). To overcome a limitation that a conventional statistical method cannot capture multidimensional aspects of temporal dietary patterns (e.g., 24-dimensional feature vectors, multivariate dietary intake time-series data), we will use a novel approach combining nutrition and systems science?machine learning method. The Interactive Diet and Activity Tracking in AARP (IDATA) study that repeatedly collected diet, anthropometry, and blood samples from 1,021 men and women, 50-74 years old will be used. During one year, the IDATA study collected 24-hour recalls with clock time for each eating occasion, every other month (total six 24-hour recalls); measured anthropometry three times (baseline and at month 6 and 12); and collected blood twice, 6-month apart. Successful completion of our proposed study will identify temporal dietary patterns that are related to diet quality and metabolic health and validate the utility of temporal dietary patterns as a new tool for future research on diet-health relations and prevention of chronic diseases.